News & Media

Latest Stories

239 results found
Filtered by Artificial Intelligence, Computational Mathematics & Statistics, Fossil Energy, Grid Architecture, Neutrino Physics, Precision Materials by Design, Subsurface Energy Systems, Vehicle Technologies, Weapons of Mass Effect, and Wind Energy
SEPTEMBER 14, 2020
Web Feature

VOLTTRON™ Goes to School

The PNNL-developed VOLTTRON™ software platform’s advancement has benefited from a community-driven approach. The technology has been used in buildings nationwide, including most recently on a university campus.

Secretary of Energy Advisory Board (SEAB) Report Recognizes PNNL Contributions

ML and AI

Report features how PNNL’s computing capabilities are affecting the nation’s security, science, and energy missions

August 25, 2020
August 25, 2020
Highlight

Contributions from researchers across Pacific Northwest National Laboratory (PNNL) were recognized in the preliminary findings of a Secretary of Energy Advisory Board (SEAB) report from a working group dedicated to the U.S. Department of Energy’s (DOE’s) capabilities and future in artificial intelligence (AI) and machine learning. PNNL researchers’ expertise is prominent throughout DOE’s AI efforts, particularly in the areas of data sciences and national security.

Based largely on input from DOE sponsors, the report features how PNNL’s computing capabilities are affecting the nation’s security, science, and energy missions. Key highlights include:

  • Studying how AI affects the global landscape for securing nuclear materials, potentially using deep learning to enhance physical and digital protections against material concealment, delivery, theft, and sabotage.
  • Describing how the United States and its partners might employ deep learning to combat attack efforts for enhanced nuclear security.
  • Designing advanced deep learning models to characterize operations with buildings, using electrical signatures on power lines, enabling new designs for energy-efficient buildings in addition to enhanced security features for nuclear facilities.
  • Leading the nuclear explosive monitoring project with data scientists working to significantly lower detection thresholds of low-yield, evasive underground nuclear explosions without increasing time-to-detection or the amount of human analysis.
  • Co-design of advanced accelerator, memory and data movement concepts to support convergence of AI and machine learning methods with other forms of data analytics and traditional scientific high performance computing (HPC). 

The report highlights PNNL’s support to the National Nuclear Security Administration, featuring joint laboratory collaborations between PNNL and others, including the Y-12 National Security Complex, Sandia National Laboratories, Lawrence Livermore National Laboratory, Los Alamos National Laboratory, and Oak Ridge National Laboratory. Additionally, PNNL is working as part of DOE’s comparative advantages in AI, providing the Office of Energy Efficiency and Renewable Energy access to AI subject matter experts.

View full preliminary findings of the Secretary of Energy Advisory Board (SEAB) report.

For more information about PNNL’s research contributions, contact Aaron Luttman

Study Shows Coastal Wetlands Aid in Carbon Sequestration

data collection in marsh

PNNL scientist, Amy Borde collects data in a marsh on the Columbia River estuary.

Photo: Heida Diefenderfer

Sea-level rise impacts will likely decrease ecosystem carbon stocks

August 13, 2020
August 13, 2020
Highlight

Tidal marshes, seagrass beds, and tidal forests are exceptional at absorbing and storing carbon. They are referred to as total ecosystem carbon stocks, yet little data exists quantifying how much carbon is absorbed and stored by tidal wetlands in the Pacific Northwest (PNW). Knowing this information is valuable, particularly in the context of sea level rise and with the associated need for Earth system modeling to predict changes at the coast.

The Science

Researchers found that the average total ecosystem carbon stock in the PNW is higher than in other areas of the U.S. and other parts of the world. Marsh carbon stocks, in particular, are twice the global average. Researchers found progressive increases in total ecosystem carbon stocks along the elevation gradient of coastal wetland types common in the PNW: seagrass, low marshes, high marshes, and tidal forests. Total carbon also increased along the salinity gradient, with more carbon occurring in lower salinity areas.

Additionally, this research showed that common methods used to estimate soil carbon actually underestimate soil carbon stocks in coastal wetlands. Soil carbon storage below the depth of 100 centimeters proved to be an important carbon pool in PNW tidal wetlands.

The Impact

The results suggest that long-term sea-level rise impacts, such as tidal inundation and increased soil salinity, will likely decrease ecosystem carbon stocks. This is a concern if wetlands can’t migrate with increased sea level due to being bound by topography and human development.  

Summary

This research arose from the Pacific Northwest Blue Carbon Working Group, of which Amy Borde and Heida Diefenderfer of Pacific Northwest National Laboratory’s Coastal Sciences Division are members. The team studied 28 tidal ecosystems across the PNW coast, from Humboldt Bay, California, to Padilla Bay, Washington. They sampled common coastal wetland types that occur along broad gradients of elevation, salinity, and tidal influences, collecting the data necessary to calculate total carbon stocks in both above ground biomass and the soil profile.

In three years of study, the researchers found that most carbon is in the wetland soils not aboveground, and much of it is deeper than one meter—a typical lower limit of sampling. Total ecosystem carbon stocks progressively increased along the terrestrial-aquatic gradient of coastal wetland ecosystems common in the temperate zone including seagrass, low marshes, high marshes, and tidal forests. The findings were reported in “Total Ecosystem Carbon Stocks at the Marine-Terrestrial Interface: Blue Carbon of the Pacific Northwest Coast, USA,” published in the August 2020 online edition of Global Change Biology (DOI: 10.1111/gcb.15248).

Research Team: PNNL’s Amy Borde and Heida Diefenderfer, along with J. Boone Kauffman, Leila Giovanonni, James Kelly, Nicholas Dunstan, and Christopher Janousek (Oregon State University); Craig Cornu and Laura Brophy (Institute for Applied Ecology/Estuary Technical Group); and Jude Apple (Padilla Bay National Estuarine Research Reserve).

Funding

The grant award was administered by the Institute of Applied Ecology, and other partners included Oregon State University and the Padilla Bay National Estuarine Research Reserve. This research was supported by the National Oceanic and Atmospheric Administration, through a cooperative agreement with the University of Michigan. 

10.1111/gcb.15248

Kauffman, J Boone, Leila Giovanonni, James Kelly, Nicholas Dunstan, Amy Borde, Heida Diefenderfer, Craig Cornu, Christopher Janousek, Jude Apple, and Laura Brophy. “Total Ecosystem Carbon Stocks at the Marine‐terrestrial Interface: Blue Carbon of the Pacific Northwest Coast, United States.” Global change biology, no. 0 (August 11, 2020). DOI: 10.1111/GCB.15248

August 11, 2020

Deep learning to generate in silico chemical property libraries and candidate molecules for small molecule identification in complex samples

January 21, 2020
May 20, 2020
Journal Article

Abstract

Comprehensive and unambiguous identification of small molecules in complex samples will revolutionize our understanding of the role of metabolites in biological systems. Existing and emerging technologies have enabled measurement of chemical properties of molecules in complex mixtures and, in concert, are sensitive enough to resolve even stereoisomers. Despite these experimental advances, small molecule identification is inhibited by (i) chemical reference libraries (e.g. mass spectra, collision cross section, and other measurable property libraries) representing <1% of known molecules, limiting the number of possible identifications, and (ii) the lack of a method to generate candidate matches directly from experimental features (i.e. without a library). To this end, we developed a variational autoencoder (VAE) to learn a continuous numerical, or latent, representation of molecular structure to expand reference libraries for small molecule identification. We extended the VAE to include a chemical property decoder, trained as a multitask network, in order to shape the latent representation such that it assembles according to desired chemical properties. The approach is unique in its application to metabolomics and small molecule identification, with its focus on properties that can be obtained from experimental measurements (m/z, CCS) paired with its training paradigm, which involved a cascade of transfer learning iterations. First, molecular representation is learned from a large dataset of structures with m/z labels. Next, in silico property values are used to continue training, as experimental property data is limited. Finally, the network is further refined by being trained with the experimental data. This allows the network to learn as much as possible at each stage, enabling success with progressively smaller datasets without overfitting. Once trained, the network can be used to predict chemical properties directly from structure, as well as generate candidate structures with desired chemical properties. Our approach is orders of magnitude faster than first-principles simulation for CCS property prediction. Additionally, the ability to generate novel molecules along manifolds, defined by chemical property analogues, positions DarkChem as highly useful in a number of application areas, including metabolomics and small molecule identification, drug discovery and design, chemical forensics, and beyond.

10.1021/acs.analchem.9b02348

Research topics

Citation

Colby S.M., J. Nunez, N.O. Hodas, C.D. Corley, and R.S. Renslow. 2020. "Deep learning to generate in silico chemical property libraries and candidate molecules for small molecule identification in complex samples." Analytical Chemistry 92, no. 2:1720-1729. PNNL-SA-144150. doi:10.1021/acs.analchem.9b02348

Distributed Wind Representation in Modeling and Simulation Tools

May 13, 2020
May 13, 2020
Report

Pacific Northwest National Laboratory (PNNL) compiled, characterized, and evaluated the inclusion of distribution wind in a number of modeling and simulation tools for the U.S. Department of Energy (DOE) Microgrids, Infrastructure Resilience, and Advanced Controls Launchpad (MIRACL) project. This work also benefits the international community participating in International Energy Agency (IEA) Wind Task 41: Enabling Wind to Contribute to a Distributed Energy Future. Tools were assessed based on a survey of publicly available and readily accessible information. The evaluation approach, methodology, key takeaways, and next steps are presented below, and the evaluation table is attached to this PDF.  

Research topics